Thermodynamic Networks: Harnessing Non-Equilibrium Steady States for Computation
This work provides a new paradigm for physics-based computation, potentially enabling low-power, autonomous computing for machine learning tasks.
The authors introduce thermodynamic networks that use non-equilibrium steady states for computation, identifying negative differential conductance as key for universal function approximation. They demonstrate high performance on sine function approximation and MNIST digit classification using quantum dot and enzymatic reaction networks.
We introduce thermodynamic networks, a general framework for autonomous, physics-based computation using non-equilibrium steady states. These networks are modeled as a collection of finite-size reservoirs that exchange conserved quantities--such as electric charge or molecular number--while relaxing to a non-equilibrium steady state, which encodes the solution of a computational problem. We identify Negative Differential Conductance (NDC) as the critical physical property governing the computational expressivity of the thermodynamic network. While networks lacking NDC are restricted to computing monotonic functions, the presence of NDC enables universal function approximation. For the training of the network, we use protocols that take advantage of the natural tendency of the system to equilibrate. We illustrate the versatility of our approach via two different platforms: quantum dot networks and enzymatic reaction networks. Both systems can be engineered to have NDC, enabling high performance in standard benchmarks, including sine function approximation and MNIST digit classification. Overall, our work establishes a rigorous link between non-equilibrium steady states and computational expressivity.